A classical twin study containing MZ and DZ twin pairs that have been raised together allows estimation of A, C or D, and E. C and D cannot be determined simultaneously because these terms are negatively confounded with each other: dominance effects increase the correlation between MZ twins, while shared environment factors increase the correlation between DZ twins (Anderson et al. 2002). Interaction terms such as gene-environment correlation can be estimated with more complex models including populations such as children of twins, an extension made possible through the use of more sophisticated techniques such as structural equation modeling (SEM), and its visual analogue, path analysis (see Fig. 4). While heritability can be estimated using simple equations based on correlations, SEM allows the creation of broader models including other family members such as siblings, or estimating the effects of variance components with factors such as age. SEM also makes it possible to systematically test whether age or other specific variance components are contributing significantly to improve the fit of the overall model (Neale and Maes 2005).
